There are two methods µPad can use to write data to the target’s non-volatile memory. It can write the data directly to the memory, or it can write the data, first into volatile memory, and then run a program (called the bounce code) on the target which copies that data to non-volatile memory. The first method is simpler but slower. The second method is generally faster, but under certain circumstances can fail. Since it entails a small overhead it might also be slower for programming small files.
By default, µPad uses the bounce method. If you want to use the direct method instead, pass the option ‘--no-bounce’.